Insect. Soc. 51 (2004) 37-42 0020-1812/04/010037-06 I Insectes Sociaux DOI 10.1007/s00040-003-0723-z © Birkhäuser Verlag, Basel, 2004 Research article The evolution of social parasitism in Acromyrmex leaf-cutting ants: a test of Emery's rule S. Sumner''^, D. K. Aanen', J. Delabie^ and J. J. Boomsma' ' Department of Population Ecology, Zoological Institute, University of Copenhagen, Universitetsparken 15, Copenhagen 2100, Denmark ^ Current address: Smithsonian Tropical Research Institute, Gamboa PO Box 2072, Balboa, Panama, e-mail: [email protected] ^ U.P.A. Laboratorio de Mirmecologia, Convenio UESC/CEPLAC, CE 7, 45600-000 Itabuna, Bahia, Brazil Received 18 November 2002; revised 16 July 2003; accepted 24 July 2003. Summary. Emery's rule predicts that social parasites and implications on the modes of evolution (Bourke and Franks, their hosts share common ancestry and are therefore likely to 1991). Social parasites may be their host's closest relative be close relatives. Within the leaf-cutting ant genus Acro- (the strict version of Emery's rule) and are likely to have myrmex, two taxa of social parasites have been found, which evolved intraspecifically from their host's lineage by sym- are thought to occupy opposite grades of permanent social patric speciation (Bourke and Franks, 1991) or through a parasitism, based on their contrasting morphologies: Acro- combination of allopatry and secondary sympatry (Lowe et myrmex insinuator differs little in morphology from its free- al., 2002). Alternatively, social parasites may be close rela- living congeneric host species and produces a worker caste, tives of their host but not sister species (the loose version of and is thus thought to represent an early grade of social par- Emery's rule). In this case the social parasite may have asitism. At the other extreme, Pseudoatta spp. exhibit a very arisen in two different ways, either interspecifically from a specialised morphology and lack a worker caste, both of non-parasitic species in a different lineage to their host which are characteristics of an evolutionarily derived grade (Bourke and Franks, 1991), or intraspecifically, from the of social parasitism. Here we present a molecular phylogeny same lineage as their host, after which the parasite may using partial sequences of cytochrome oxidase I and II of have switched hosts or imdergone subsequent speciation about half of the known Acromyrmex species including two (Buschinger, 1990). The problem with testing Emery's rule social parasites, their hosts and all congeneric species occur- through phylogenetic analysis therefore is that it is difficult ring sympatrically. We show that the two inquiline parasites to distinguish between events that occurred at the origin of represent two separate origins of social parasitism in the social parasitism and events that occurred afterwards. The genus Acromyrmex. The early-grade social parasite A. insin- highest probability of tracing events at the moment of spe- uator is highly likely to be the sister species of its host ciation will be for species close to the speciation event, i. e. Acromyrmex echinator, but the derived social parasite incipient parasites, which are unlikely to have switched from Pseudoatta sp. is not the sister species of its extant host their original host or undergone secondary speciation. Such Acromyrmex rugosus. comparisons are particularly useful if they can be contrasted with closely related lineages of advanced social parasitism Key words: Inquiline parasites, speciation, phylogeny, Acro- for which the strict version of Emery's rule is least likely to myrmex. apply Of the few phylogenetic analyses conducted on hymen- opteran social parasites and their hosts, most support at least the loose form of Emery's rule. A rare exception exists in Introduction which host and parasite belong to different subfamilies (Maschwitz et al, 2000). Social parasitism in bees (Dan- Emery's rule states that social parasites are close relatives of forth, 1999; Pedersen, 1996) and in wasps (Carpenter et al, their hosts (Emery, 1909) and thus allows specific predic- 1993; Choudhary et al, 1994) tends to be concentrated in tions about the ancestry of social parasites and inferences relatively few species-rich clades of monophyletic origin, about their evolution and modes of speciation to be made. which often exploit species of equally extensive sister clades. Emery's rule can be resolved at two levels and it is important Many genera of ant social parasites are polyphyletic and to distinguish between them because they have different often show highly convergent adaptations (e.g. Ward, 1989; 38 S. Sumner et al. Social parasites in leaf-cutting ants Agosti, 1994; Baur et al., 1996; Ward, 1996; Savolainen and genus Acromyrmex. Our specific aim was to test the degree to which Vepsäläinen, 2003). An exception is two Pogonomyrmex Emery's rule is supported and from this make inferences about likely inquiline parasites which form a monophyletic group within modes of speciation of socially parasitic lineages within Acromyrmex. Our selection of species reflects this objective: we included the repre- the host clade (Parker and Rissing, 2002). Nonetheless, it has sentatives of the two socially parasitic lineages, their extant host species not been easy to assess if and when Emery's rule holds strict- and all other Acromyrmex species occurring sympatrically with these ly amongst ant social parasites and their hosts because few parasites. In addition, we included a random selection of other species host-parasite pairs have been subjected to molecular phylo- from Central and South America that were made available to us (see genetic analysis (Ward, 1989). Furthermore, almost no stud- table 1 for collection sites and species ranges). We combined our own sequence data with two Acromyrmex sequences {A. octospinosus and ies have considered incipient parasites, because they are rare. A. volcanus), three Atta sequences {Atta sexdens sexdens, Atta sexdens The reported cases amongst the ants where Emery's rule does rubropilosa and Atta cephalotes) and two non-leafcutting attines (Tra- seem to hold strictly are based on morphological evidence chymyrmex saussurei and Sericomyrmex amahilis) whose comparable (Elmes, 1978; Wilson, 1984; Schultz et al., 1998, but see sequences were available in Genbank (Wetterer et al., 1998). The latter Savolainen and Vepsäläinen, 2003). two were used as outgroups. Genomic DNA was extracted from the thorax of 1 -4 individuals The incipient inquiline parasite Acromyrmex insinuator for each species using standard techniques. A mitochondrial DNA frag- was discovered recently (Schultz et al., 1998). It belongs to ment was amplified using the polymerase chain reaction (PCR). The the leaf-cutting ants, Attini (Myrmicinae) and is found in fragment included the 3' part of COI, an intergenic spacer, the tRNA Panama, Central America where it parasitizes a single host leucine locus and the 5' end of COIL In total the fragment amplified was Acromyrmex echinatior. It is thought to be incipient because approximately 600 base pairs. These partial genes were amplified in each species using primers designed from existing leaf-cutting ant it closely resembles its host in morphology and produces a sequences (Wetterer et al., 1998). The primers used for each species are worker caste (Schultz et al., 1998). Morphological data sug- indicated in Table 1. Approximately 50 ng of template DNA were ampli- gest that A. insinuator is the sister species of its host, in fied in 50 \A volumes in 1 x reaction buffer, 0.2 mM dNTPs, 0.5 jiM agreement with the strict version of Emery's rule (Schultz et primers, with 1.25 U of Taq polymerase, using an initial denaturing step al, 1998; Bekkevold and Boomsma, 2000). Because few of 94°C for 2 min, followed by 25 cycles of 94°C for 30 s, 48-55°C for 30 s (depending on species and primer pair), 72 °C for 30 s. A final elon- incipient social parasites with a worker caste are known, gation step of 72°C for 10 min completed the amplification process. A. insinuator presents a rare opportunity to examine the true Products were run on 2% agarose gels stained with ethidium bromide. origin of inquiline behaviour (Sumner et al., 2003). Another Single strong bands were cleaned up using the Qiagen kit and subse- two inquiline parasites of Acromyrmex leaf-cutting ants have quently sequenced. been found in South America. Together they presently con- All 19 sequences were aligned using the program Sequencher 3.11 (Gene Codes Corporation). Phylogenetic relationships were inferred stitute their own genus, Pseudoatta (Pseudoatta argentina, from the aligned sequences using unweighted parsimony with PAUP * host, Acromyrmex lundi (Santschi, 1926; Gallardo, 1929) 4.0b8 and 10 (Swofford, 2001). Gaps were coded as missing data. The and Pseudoatta sp., host, Acromyrmex rugosus (Delabie et Branch and Bound option was used in PAUP to find the most parsimo- al, 1993)). These two inquiline parasites represent an nious trees. Clade stability was assessed by 1000 bootstrap replications extreme grade of social parasitism where the worker caste (Hillis and Bull, 1993), using heuristic searches to find the best trees in has been lost and the morphology has become specialised each replicate (a stepwise addition starting tree and 10 random addition sequences with TBR branch swapping). In addition to the parsimony and very different from that of their hosts. analysis, three separate Bayesian analyses were performed using Twenty six species of the genus Acromyrmex have been MrBayes 3.04b (Huelsenbeck and Ronquist, 2001, http://morphbank. described from throughout the Neotropics (HöUdobler and ebc.uu.se/mrbayes/info.php). MrModeltest 1,1b * (Nylander, 2002) was Wilson, 1990 supplemented with recent literature). Two phy- used to select the best-fit model. The best-fit model was the general logenies based on larval morphology and molecular se- time reversible model of sequence evolution with gamma and a propor- tion of invariable sites (GTR + gamma + I; base frequencies estimated: quence data have so far been constructed for the attine ants.